Protein antigenic determinants have been classified as continuous or discontinuous. The continuous determinants are composed of residues which are local in the polypeptide sequence, while discontinuous determinants consist of residues from different parts of the sequence, brought together by the folding of the protein to its native structure. Searches made for protein determinants using peptide fragments which compete with protein-antibody complex formation, or peptides that can be used to raise antibodies which crossreact with the native protein, are limited to the simulation of continuous determinants. However, recent experiments suggest that most determinants are discontinuous. We now show, by consideration of protein surfaces, that if the recognition zone between a protein and antibody has the same dimensions as those found for the lysozyme-antibody complex, none of the protein's surface will be 'continuous'. We suggest that all determinants are discontinuous to some extent, and that crossreacting peptides mimic only the 'primary' interaction site. In addition, we show that the parts of a protein's surface which are most continuous fall predominantly in the loops and/or protruding regions. This explains why quantities such as hydrophilicity, accessibility, mobility and protrusion can be used to predict which parts of a polypeptide provide the 'best' antigenic peptides.